The availability of dialogue context significantly influences the quality and consistency of crowdsourced evaluation labels for task-oriented dialogue systems.
